Let us re-think the Cloud strategy

Businesses may find themselves trapped in the cloud and soon realise that cloud-hopping is not exactly like pub-hopping or changing lanes

Watching ‘The Menu’ movie feels eerie from the very first scene. Despite the red-carpet welcome, the meticulous arrangements, the air of exclusivity, the suave staff, and the company of high society, something is off in that elegant island. The guests are supposed to be treated to a specially crafted many-course haute cuisine by a coveted chef. And amidst well-marinated satire, dark humour, and mind-shaking epiphanies, what follows is unexpected turbulence, knives not meant for slicing food, blood gushing out between fancily-plated gourmet plates and a suffocating sense of, well, being locked in.

“Most cloud providers structure their services as roach motels to make it easy to move in and difficult to leave.”- Mike Loukides, Vice President, Content Strategy, O’Reilly Media.

Guests try to run, to break windows, to scream, to throw tantrums – but nothing helps. Unless a girl, who was never supposed to be on this special list, does something obvious and unexpected. And she gets to get out. Exactly how? We will come to that in a bit.

But it is an attempt that some enterprises also seem to opt for due to indigestion from the Cloud buffet they opted for. They are trying everything: repatriation, hybridisation, scale-back, right-sizing, moving to special-purpose hardware or infrastructure optimisation. It can be due to unexpected shocks of Cloud economics, hidden Cloud bills that were not visible before, indirect breathlessness due to vendor lock-in or some data control and sovereignty issue.

When does it feel ‘Not Right’

Let us start with the why. And what better way than to read what David Heinemeier Hansson, Co-founder and CTO of 37signals wrote in a 2022 post? In “Why we are leaving the Cloud.” he argued that Basecamp had one foot in the Cloud for well over a decade, and Hey has been running there exclusively since it was launched two years ago.

“We have run extensively in both Amazon and Google Cloud. We have run on bare virtual machines, and we have run on Kubernetes. We have seen all the Cloud has to offer and tried most of it. It is finally time to conclude: renting computers is (mostly) a bad deal for medium-sized companies like ours with stable growth. The savings promised in reduced complexity never materialised. So, we are making our plans to leave.”

According to his bean-counting, continuing to operate in the Cloud, is like paying an absurd premium for the possibility that something could go wrong. “It is like paying a quarter of your house’s value for earthquake insurance when you do not live anywhere near a fault line. Yeah, sure, if somehow a quake two states over opens the earth so wide it cracks your foundation, you might be happy to have it, but it does not feel proportional, does it?”

This echoes what Dropbox did when it announced weaning itself off from the Cloud and garnering millions of dollars in savings (about USD 75 million reportedly in 2017) in that shift – by repatriating workloads from the public Cloud.

As Hansson decided, “We consider it a duty that we at 37signals do our part to swim against the stream. Our business model is incredibly compatible with owning hardware and writing it off over many years. Growth trajectories that are mostly predictable. Expert staff who might as well employ their talents operating our machines as those belonging to Amazon or Google. And I think there are plenty of other companies in similar boats.”

There are examples of a hybrid approach too, as seen with CrowdStrike and Zscaler. According to a 16z analysis, across 50 of the top public software companies currently utilising Cloud infrastructure, as much as an estimated USD 100 billion of market value is being lost among them due to Cloud impact on margins — relative to running the infrastructure themselves. When the analysis is extended to the broader universe of scale public companies – that stand to benefit from related savings – the total impact is reckoned to be potentially greater than USD 500 billion.

The Andreessen Horowitz analysis also pointed out: “For every dollar of gross profit saved, market caps rise on average 24-25X the net cost savings from Cloud repatriation (assumed savings are expressed net of depreciation costs incurred from incremental Capex if relevant). This means an additional USD 4 billion of gross profit can be estimated to yield an additional USD 100 billion of market capitalisation among these 50 companies alone. Repatriation results in one-third to one-half the cost of running equivalent workloads in the Cloud.” Interestingly, Cloud repatriation is estimated to drive a 50 per cent drop in Cloud spending, translating into total savings of USD 4 billion in recovered profit.

“We expect to see a rise in demand for native cloud environments to manage systems efficiently through reduced dependence on traditional servers.”- Manish Gupta, Vice President & General Manager, Infrastructure Solutions Group, Dell Technologies India

“Humans are wired to be in control. We are emotional beings. And an on-premise system is no different – where you and your team can see and fix everything in proximity.”- Indranil Bandyopadhyay, Principal Analyst, Forrester

A lot of CIOs prefer either an on-premise or hybrid approach. As Bhoopendra Solanki, Chief Information Officer, Sakra World Hospital says, “The data size and compliance factors are also a reason here. In industries like BFSI and healthcare, data sensitivity is on the higher side. No matter how much assurance a Cloud provider gives, core areas cannot be completely on Cloud. We have a hybrid approach here.”

Speaking to Indranil Bandyopadhyay, Principal Analyst, Forrester gives us a chance to look at this subject from both an incisive and a hands-on view. His take, in short, is that Cloud may not be a panacea for core systems for large enterprises with heritage systems. “It can be a good idea for young and Cloud-native companies for the Opex advantage. It can also be a good option for tertiary use cases where heavy data processing can be handled in the Cloud. It will also be a plausible move for stuff that enterprises want to innovate with low-code and Cloud infra.” He, however, cautions that organisations must think again for anything where the business is in a direct impact area, regulators have data-location concerns, and they cannot afford damage if they are picking Cloud just because everyone else is.

Here is why, as he explains in detail.

Courses for horses

“In many companies with heritage technology, like the Mainframes or the AS400s, there are core systems that are not engineered for Cloud models. People do not want to fix something which is not broken. Especially when the technology is well-entrenched, working fine and is pretty fault-tolerant. If one moves to Cloud in such a back-drop, it will take a lot of re-purposing and system modernization,” Bandyopadhyay dissects.

Looking back at his days as a practitioner, he cites an example to explain this. “When our business made a strategic decision to move to Cloud, I looked at all the hardware and realised how hardware and software are financially different. Hardware gets depreciated quicker than software, maybe in just 2-3 years. And in an on-premise scenario, the Opex is explicit, but in Cloud, the Opex is implicit. The idea of variable costs that Cloud vendors sell is amazing, but it may not work if the hardware is sitting idle and has an accounting connotation for me. If the hardware is doing its job, then what is my financial reason to move to Cloud? It is also not a lift-and-shift choice, as many perceive it to be. To take advantage of Cloud and variable costs, one has to do a lot of refactoring.”

To add to that there are other complications like migration of data. The bundled services offered by some Cloud players can turn out to be more expensive than expected, he reminds.

Mike Loukides, Vice President, Content Strategy, O’Reilly Media weighs in that Cloud switching and roll-back is a palpable and stable trend. “There are two separate trends here. Cloud rollback is often called ‘Cloud repatriation’ which is taking applications and data out of the Cloud and moving them back to company datacentres. Cloud switching is moving from one provider to the other.” Loukides explains that in both cases, Cloud users will do what is economically necessary. “Many companies are finding the ‘Cloud’ to be more expensive than they initially thought and therefore moving back to on-premise or hosted solutions.”

Loukides recommends that it is worth asking whether their Cloud expenses are high because their savings estimates were over-optimistic, or because they used the Cloud inefficiently: spinning up more instances than they needed, not shutting down instances that were not needed, and/or not taking advantage of discounted services like AWS Spot Instances. “IT managers need to look at this carefully before deciding to pull the plug on the Cloud. Are you using the Cloud efficiently, in a way that minimises costs? It would be a shame to repatriate if the real reason for repatriation is that you have not used the Cloud effectively.”

Manish Gupta, Vice President and General Manager, Infrastructure Solutions Group, Dell Technologies India captures that in the current times, Cloud is essential in storing the data while businesses adapt to the new digital goals. “The shift to the Cloud is going to be pivotal for Indian enterprises to be successful in the digital era. While businesses try to venture out and combine services to save on expenditure components, the changes in the IT landscape will push businesses to explore the Cloud.”

“Globally, and in India, we expect to see investments in hybrid Cloud operating models that span public, private and edge environments to grow to enable rapid scale and management of IT workloads. A multi-Cloud approach will define processes of the future to store and manage mission-critical data. With time, CIOs across organisations shall realise that the smaller ongoing costs versus cyclic infrastructure builds, will bring more value to the business and change perceptions of IT, being considered as a cost-heavy component to the business.”

In Gupta’s opinion, private Clouds also provide greater levels of security and control, while the decentralised processing of Cloud edge computing can help to reduce costs and enable low-latency experiences on edge devices. For example, Dell Technologies’ Cloud simplifies its migration by providing IT teams with a consistent management experience and familiar tools, creating efficiencies and increasing comfort with tasks and reducing errors.

How to leave the table?

Moving from one Cloud to another is a different scenario, admits Loukides. “Most Cloud providers structure their services as ‘roach motels’ to make it easy to move in and difficult to leave. Dropping one Cloud for another is likely to be expensive and may not yield any significant savings. However, all Cloud providers are not the same. In some cases, you may need a service that another provider offers. In this case, a better alternative to Cloud switching would be designing the application to run across multiple providers: splitting the application into parts, each of which can run whichever provider delivers the service it needs.”

Bandyopadhyay opines that it is not always such an easy choice, especially as Economics has changed during Covid and post-Covid world, and when companies have become money-conscious. Transferring anything from on-premise to Cloud also means a lot of disruptions; not just qualitatively, but also quantitatively.

Examining how easy and practical it is for someone to switch back to the on-premise model, Bandyopadhyay underlines it all in two words: courage and deep pockets.

“If a company has the culture of accepting a mistake of ‘sunk costs’ and correcting it, it will make the switch. But it has to have deep pockets to do that. Because just like going to Cloud needed a lot of work, going back from Cloud would need a lot of re-engineering too. And that means a lot of time.” Here the companies that have suffered any real business damage due to a Cloud investment are more likely to hit the ‘reset’ button, Bandyopadhyay weighs in.

Would SLAs be a way to accommodate these expectations and create a safety net for enterprises? Bandyopadhyay reasons that while claw-back mechanisms can be put in place, the question is not about money; it is about damage. “Does it help to be paid by a Cloud vendor if you are a bank and you suffer Cloud outage? The claw-back can put some pressure on vendors but it is not a silver bullet for your problems.”

It is not that easy to do Cloud-switching. Gupta avers that moving workload between public, private and edge Clouds can be expensive. “It is also time-consuming and risky. Cloud migration can take months without consistent management infrastructure, adversely impacting productivity and hindering innovation. The major challenge is the difficulty that IT teams experience when working with Cloud resources that have drastically different management frameworks, tools, SLAs and security issues.”

So, what is the right way about it? “It is not always about choosing a private or a public Cloud solution, the decisive move has to be about aligning the needs of the on-prem infrastructure with the expectations from the Cloud environment of the organisation. We could have a datacentre but run applications with multiple hyperscalers, hence without a consistent infrastructure between Clouds, the task of migration can take much longer and place the organisation at greater risk. Ease of access to data, proactive security and making the value chain resilient to threats, and the best agility, flexibility and scalability for business needs, would determine what Cloud infrastructure an organisation would invest in. Latency is also a determining factor in choosing the kind of Cloud for the business, be it public, private or even, edge,” says Gupta.

Of the top public software companies currently utilising Cloud infrastructure, an estimated USD 100 billion of market value is being lost among them due to cloud impact on margins.

Not one fork, but many

Loukides reflects how a decade ago, it was hard for IT groups to think about architectures like this. “Now, it is common with very few companies using a single Cloud. It may have a Cloud strategy built around AWS, Azure or GCP, but eventually, there will be a wildcat project running on a different provider that suddenly becomes important; or there will be an acquisition, and the acquired company will have its applications running in another Cloud. Everyone is (or will be) multi-Cloud, and that is a good thing.”

Gupta argues that a shift to an on-demand service offering for hybrid Cloud computing will provide organisations with an efficient Cloud environment that can develop, manage and deliver applications with reduced spending in the longer run. “That being said, consistent operations and infrastructure across Clouds are paramount. Indian businesses must realise that the hybrid Cloud model is the right strategy for longer-term costs, scalability and security. At Dell Technologies, through the VMware Cloud, we are already offering Infrastructure-as-a-Service services to help enterprises with the ability to move workloads across multiple Cloud environments and scale resources quickly with predictable pricing and transparent costs.”

When it comes to Opex-heavy businesses or young and Cloud-born companies, it can be a good choice to stick to Cloud, reasons Bandyopadhyay. “For them, Opex is a big barrier to entry and Cloud creates a level-playing field. Also, if you are a company dealing with AL, ML, deep learning models or some form of heavy batch-processing or quantum workloads, it will make sense to take some things to Cloud.”

However, Europe has not seen much Cloud uptick due to the data location concerns; the regulators are asking hyperscalers if there is enough care taken for data scrutiny by other governments. “Of course, there are ways to solve that with encryption layers, but regulators decide based on the Lowest Common Denominator. You cannot ignore the instances of Cloud-setups going down, and in such cases, a business cannot tell its customers that it is because of a Cloud outage. You are still answerable for your business data and uptime.”

Bandyopadhyay sums it up well. “Cloud may not tick all the boxes. Go with your eyes open. Do not think of it as some Panacea. Go for Cloud for the right reason. Not because it is being painted as the future. Even if you want Cloud for its agility and as an additional capacity and low-code buddy to help you with experimentation, It is not an ‘or’ decision but an ‘and’ decision.”

He has a beautiful philosophical angle to explain the inexplicable here. “Human beings love control. I may be okay outsourcing the care of my car to someone better equipped to deal with it, but I still want to see my wife, my kids, and my pets. I cannot give that away. Humans are wired to be in control. We are emotional beings. And an on-premise system is not different – where you and your team can see and fix everything in proximity.”

Your time to clap

Cloud, has for a long, been promised and proven a great choice for companies on a growth inflection, and with a need for high elasticity of computing in unpredictable demand scenarios. And while disillusionment with Cloud can be due to gaps that Cloud providers have failed to address or due to external factors like security and outages or availability of cheaper hardware outside, one cannot deny that it may be easy to get into an island full of promises, but it is never easy switching off a Cloud, or from a Cloud.

Unless you have the clarity of exactly what you

want and ask the ruthless, but still-human, Chef for it. With a voice that is confident when it says “I want to send your food back. I do not like it, Chef.” Like the girl Margot did – and got away from the island while it burnt down.

By Pratima H

pratimah@cybermedia.co.in